Model Price

What does it cost the user per call? Look here.

What it is

ModelPrice is the system's "price list". Each model has one row, deciding:

• Token billing: ¥ per million input / output tokens
• Per-request billing: ¥ per call (used for DALL·E / image generation etc.)

Per-call deduction = consumption × ModelPrice × group discount.

Where to find it

• Admin → Model Prices: list, add, bulk edit
• Admin → Dashboard: today's / this month's revenue per model

What to set

Admin → Model Prices → Add:

Field	Meaning	Effect
model_name	Model name	Must match a value in Channel → Models; mismatched names can't be called
billing_type	token (per token) or per_request (per call)	Decides the four fields below
input_price	¥ per million input tokens	Drives input-side deduction
output_price	¥ per million output tokens	Drives output-side deduction
cached_price	¥ per million cached tokens	Usually half of input_price; 0 = no cache
per_request_price	¥ per call	Only used when billing_type=per_request
enabled	Toggle	Disabled → model can't be called

The system seeds default prices for common models (GPT-4o / Claude / DeepSeek etc.) on first start. Edit them freely.

FAQ

• New model but not in the channel dropdown: add the model here with enabled=true first.
• Call succeeded but no quota deducted: check the model is in Model Price and the price isn't zero.
• Changed price — how is it billed going forward: calls use the model's current ModelPrice at the time of the call.

Related

• Model Price Management (admin)
• Model Price Schema
• Group Price

Related API

• GET /api/model_price/ — list (Root)
• GET /api/model_price/options — enabled model names only (Admin; for channel edit dropdown)
• POST /api/model_price/ — add
• PUT /api/model_price/ — update
• DELETE /api/model_price/:id — delete